Measurement and Verification Operational Guide

Transcription

1 Measurement and Verification Operational Guide Whole Building Applications

2 Copyright State of NSW and Office of Environment and Heritage With the exception of photographs, the State of NSW and Office of Environment and Heritage are pleased to allow this material to be reproduced in whole or in part for educational and non-commercial use, provided the meaning is unchanged and its source, publisher and authorship are acknowledged. Specific permission is required for the reproduction of photographs. The Office of Environment and Heritage (OEH) has compiled this guideline in good faith, exercising all due care and attention. No representation is made about the accuracy, completeness or suitability of the information in this publication for any particular purpose. OEH shall not be liable for any damage which may occur to any person or organisation taking action or not on the basis of this publication. Readers should seek appropriate advice when applying the information to their specific needs. Every effort has been made to ensure that the information in this document is accurate at the time of publication. However, as appropriate, readers should obtain independent advice before making any decision based on this information. Published by: Office of Environment and Heritage NSW 59 Goulburn Street, Sydney NSW 2000 PO Box A290, Sydney South NSW 1232 Phone: (02) (switchboard) Phone: (environment information and publications requests) Phone: (national parks, climate change and energy efficiency information, and publications requests) Fax: (02) TTY: (02) Website: Report pollution and environmental incidents Environment Line: (NSW only) or See also ISBN OEH 2012/0993 December 2012 Printed on environmentally sustainable paper

3 Table of contents i Contents 1 Your guide to successful M&V projects Using the M&V Operational Guide The Whole Building Applications (this guide) Understanding M&V concepts Introducing key M&V terms Best practise M&V process Getting started Proposed whole building ECM(s) Decide approach for pursuing M&V M&V design and planning M&V design Prepare M&V plan Data collection, modelling and analysis Measure baseline data Develop energy model and uncertainty Implement ECM(s) Measure post retrofit data Savings analysis and uncertainty Finish Reporting Project close and savings persistence M&V Examples Examples from the IPMVP Examples from this guide Appendix A: Example scenario A food processing plant Getting started Summary of M&V plan Baseline model development Developing the energy model for electricity Validating the regression outputs Developing the energy model for natural gas Calculating combined energy savings and uncertainty Reporting results Appendix B: Example scenario B commercial building Getting started Summary of M&V plan Baseline model development Calculating electricity savings from Model Uncertainty analysis for Model Reporting results from Model Reporting results from Model

4

5 Your guide to successful M&V projects 1 1 Your guide to successful M&V projects The Measurement and Verification (M&V) Operational Guide has been developed to help M&V practitioners, business energy savings project managers, government energy efficiency program managers and policy makers translate M&V theory into successful M&V projects. By following this guide you will be implementing the International Performance Measurement and Verification Protocol (IPMVP) across a typical M&V process. Practical tips, tools and scenario examples are provided to assist with decision making, planning, measuring, analysing and reporting outcomes. But what is M&V exactly? M&V is the process of using measurement to reliably determine actual savings for energy, demand, cost and greenhouse gases within a site by an Energy Conservation Measure (ECM). Measurements are used to verify savings, rather than applying deemed savings or theoretical engineering calculations, which are based on previous studies, manufacturerprovided information or other indirect data. Savings are determined by comparing post-retrofit performance against a business as usual forecast. Across Australia the use of M&V has been growing, driven by business and as a requirement in government funding and financing programs. M&V enables: calculation of savings for projects that have high uncertainty or highly variable characteristics verification of installed performance against manufacturer claims a verified result which can be stated with confidence and can prove return on investment demonstration of performance where a financial incentive or penalty is involved effective management of energy costs the building of robust business cases to promote successful outcomes In essence, Measurement and Verification is intended to answer the question, how can I be sure I m really saving money? Using the M&V Operational Guide The M&V Operational Guide is structured in three main parts; Process, Planning and Applications. Process Guide: The Process Guide provides guidance that is common across all M&V projects. Practitioners new to M&V should start with the Process Guide to gain an understanding of M&V theory, principles, terminology and the overall process. Planning Guide: The Planning Guide is designed to assist both new and experienced practitioners to develop a robust M&V Plan for your energy savings project, using a step-by-step process for designing a M&V project. A Microsoft Excel tool is also available to assist practitioners to capture the key components for a successful M&V Plan. Applications Guides: Seven separate application-specific guides provide new and experienced M&V practitioners with advice, considerations and examples for technologies found in typical commercial and industrial sites. The Applications Guides should be used in conjunction with the Planning Guide to understand application-specific considerations and design choices. Application Guides are available for. 1 Source: 1.1

6 2 Measurement and Verification Operational Guide Application Guides are available for: Lighting Motors, pumps and fans Commercial heating, ventilation and cooling Commercial and industrial refrigeration Boilers, steam and compressed air Whole buildings Renewables and cogeneration Figure 1: M&V Operational Guide structure 1.2 The Whole Building Applications (this guide) The Whole Building Applications Guide provides specific guidance for conducting M&V for large and/or multiple projects throughout a site. It is designed to be used in conjunction with the Process Guide, providing tips, suggestions and examples specific to conducting evaluation using a facility boundary. 1.2

7 Your guide to successful M&V projects 3 The Whole Building Applications Guide is presented as follows: Understanding M&V concepts Getting started M&V design and planning Data collection, modelling and analysis Finish References to examples of M&V projects Example whole building scenario A Example whole building scenario B Section 2 presents a high level diagram of the best practise M&V process. Section 3 provides a discussion on key things that need to be considered when getting your M&V project started. Section 4 provides guidance on how to design and plan your whole building M&V project and key considerations, potential issues and suggested approaches. Section 5 provides guidance on data collection, modelling and analysis for your whole building M&V project. Section 6 provides a discussion on reporting M&V outcomes, ongoing M&V and ensuring savings persist over time. Section 7 provides a reference list of example projects located within the IPMVP and throughout this guide. Appendix A illustrates the M&V process using a worked example of a food processing plant project Appendix B illustrates the M&V process using a worked example of a commercial building project 1.2

8 4 Measurement and Verification Operational Guide 2 Understanding M&V concepts 2.1 Introducing key M&V terms The terms listed in Table 1 below are used throughout this guide and are introduced here to assist with initial understanding. Refer to Section 4 within the Process Guide for a full definition and explanation. Table 1: Key M&V terms M&V Term Definition Examples Measurement boundary A notional boundary that defines the physical scope of a M&V project. The effects of an ECM are determined at this boundary. Whole facility, sub facility, lighting circuit, mechanical plant room, switchboard, individual plant and equipment etc. Energy use Energy used within the measurement boundary. Electricity, natural gas, LPG, transport fuels, etc Key parameters Data sources relating to energy use and independent variables that are measured or estimated which form the basis for savings calculations. Instantaneous power draw, metered energy use, efficiency, operating hours, temperature, humidity, performance output etc. M&V Options Four generic approaches for conducting M&V which are defined within the IPMVP. These are known as Options A, B, C and D. Routine adjustments Routine adjustments to energy use that are calculated based on analysis of energy use in relation to independent variables. Energy use may be routinely adjusted based on independent variables such as ambient temperature, humidity, occupancy, business hours, production levels, etc. Non routine adjustments Once-off or infrequent changes in energy use or demand that occur due to changes in static factors Energy use may be non routinely adjusted based on static factors such as changes to building size, facade, installed equipment, vacancy, etc. Unanticipated events can also temporarily or permanently affect energy use. Examples include natural events such as fire, flood, drought or other events such as equipment failure, etc. Interactive effects Changes in energy use resulting from an ECM which will occur outside our defined measurement boundary. Changes to the HVAC heat load through lighting efficiency upgrades, interactive effects on downstream systems due to changes in motor speed/pressure/flow, etc. Performance Output performance affected by the ECM. System/equipment output (e.g. compressed air), comfort conditions, production, light levels, etc. 2.1

9 Understanding M&V concepts Best practise M&V process The following figure presents the best practise M&V process which is how the rest of the Whole Building Applications Guide is structured. Refer to the Process Guide for detailed guidance on the M&V processes. Figure 2: Best practise M&V process with references to M&V Process Guide 2.2

10 6 Measurement and Verification Operational Guide 3 Getting started 3.1 Proposed whole building ECM(s) Whole building projects In previous sections we have discussed M&V for various applications with most guidance relating to Option A and B boundaries. Within this section we focus on a facility/sub-facility level boundaries using Options C or D. Whole building analysis is often preferred for the following reasons: Often utilises existing metering, energy invoices and other business data Can determine actual savings in relation to cost and usage from energy bills Enables multiple ECMs to be evaluated within a single M&V exercise which: Reduces M&V costs Avoids additionality issues Encompasses all interactive effects Useful where the baseline period would otherwise be too short as existing data covering a longer period can be used Ongoing M&V can be performed Can usually be conducted as a desktop exercise. Whole building M&V has its drawbacks, which include: Savings must be greater than 10% of usage at the project boundary (i.e. site) to be claimable Effects of individual ECMs cannot be separately evaluated Baseline period is 12 months or more, and post-retrofit period is typically 6 months or more Relies on our ability to create an energy model for the facility, which may be difficult for complex sites or those that cannot be explained in relation to independent variables Key points to note When considering an M&V, it is important to understand the nature of the site and proposed EMC(s) (what, where, when, why, how much) and the project benefits (e.g. energy, demand, greenhouse gas and cost savings). Option C Key points to note when getting started using Option C are: The proposed ECMs should be reviewed to ensure that combined savings is estimated to exceed 10% of the site s total energy use. Where multiple fuels are involved this should extend to each fuel individually. The advantage of analysing fuels individually is that different models can be created, using only those independent variables and static factors that apply. Alternatively, analysis may be conducted on total energy, which is a combination of all fuels. In addition to energy use, the identification of the right key parameters is important. Prior to modelling, all potential parameters should be identified. Data availability and quality should be assessed. Where data deficiencies exist for important variables, additional effort may be required to collect better data. This may require searching for alternate data sources, identifying a proxy, or choosing a measurement period for which suitable data is available. An Option D approach could also be considered. 3.1

11 Getting started 7 The nature of the ECMs to be implemented is only somewhat important. It is a good idea to understand the type of ECM and the areas it may affect, however with a facility level boundary it will not alter the data that is collected. Consideration should be made if the ECM involves changes to independent variables to ensure that the developed model will provide a realistic adjusted baseline once the project is implemented. Measurement periods are typically 12 months or more for baseline data and 6 months or more for post-retrofit data. Ongoing M&V may be considered, assuming that ongoing data for energy use, independent variables and static factors can be collected and that the baseline energy model is still valid. Using an ongoing approach, the effects of a staged implementation program can be progressively reported. Determine the desired level of uncertainty (precision + confidence). Determine the required and desirable M&V outcomes. Savings accuracy is dependent on the accuracy input data and the accuracy of the developed model. Option D Key points to note when getting started using Option D are: Computer simulation software is used to predict the facility energy use. This is applicable for new buildings and for retrofits where baseline data cannot be obtained. Energy use is predicted by inputting all key characteristics of the building, including: Physical properties of building materials (walls, floors, roof, windows, partitions, doors, insulation) Site specific information, including location, reflectivity of external surfaces, wind conditions, shading from external objects Infiltration rates (how well the building is sealed) Internal loads, including occupancy density, metabolic rate (activity levels and clothing) Occupancy patterns (occupation levels at various times) Equipment details (energy using equipment including size, power draw, equipment usage patterns and control strategies, heat output) Lighting details (types, numbers, locations of lights and/or lighting density, heat output, operating schedules) Air conditioning / refrigeration equipment (type and operation, electrical loads, coefficients of performance, operating schedules and control strategies) Data for independent variables and process loads (e.g. weather/temperature, wind, rainfall, process equipment and corresponding heat loads, raw material inputs, and production outputs) Determine the desired level of uncertainty (precision + confidence). Determine the required and desirable M&V outcomes. Accuracy is determined from the uncertainties associated with data collection, combined with the calibration error between the developed model and the actual energy use data. When the building is modelled, various scenarios are considered, known as Off-Axis or what if scenarios. Examples include: Part of building switches from 9am-5pm operation to 24/7. Variable air volume dampers get stuck open Tri-generation system has to be shut down. Accurate computer modelling and calibration against actual usage are key challenges. Simulation results must match for both usage and demand from monthly data. Modelling incorporates 12 months of data, usually in hourly or half-hourly increments. Ongoing M&V may be considered, assuming that ongoing data for energy use, independent variables and static factors can be collected and that the baseline energy model is still valid. 3.1

12 8 Measurement and Verification Operational Guide 3.2 Decide approach for pursuing M&V Once the nature of the M&V project is scoped and the benefits assessed, the form of the M&V can be determined. Decide which M&V approach you wish to pursue: 1. Conduct project-level M&V 2. Conduct program-level M&V using a sample based approach incorporating project level M&V supplemented with evaluation within the program population. 3. Adopt a non-m&v approach in which savings are estimated, or nothing is done. 3.2

13 M&V design and planning 9 4 M&V design and planning 4.1 M&V design M&V Option Option C is the most common form of facility based analysis, which involves the development of an energy model from historical meter/billing data combined with data for independent variables across the site. Option D is used where there is no baseline data. The site is modelled in detail, usually via specialised energy modelling software. The model is built from the bottom up, including a full asset list and equipment ratings, operating schedules, control strategies, and data for sitebased independent variables. Both options involve a long-term approach for conducting M&V Measurement boundary Either facility level boundary using Option C or Option D. Option D may also be used in conjunction with an ECM boundary. The energy system s usage must be separated from the rest of the facility using a permanent meter Key parameters For whole building M&V we need to consider the key parameters shown in the diagram and table below. Routine Adjustments Non Routine Adjustments Energy Use Equipment, Energy System, Functional Area of Whole Facility Performance M&V Option Interactive Effects 4.1

14 10 Measurement and Verification Operational Guide Table 2: key parameters to be considered when conducting M&V for whole building projects Parameter Energy use Description Energy use is typically obtained from the site s incoming revenue meter or from energy invoices. Measurement periods may be short, such as 15/30 minute interval data for electricity or natural gas, through to monthly or quarterly invoices. Where billing periods are long, meters can be read periodically (e.g. weekly, monthly) to segment billing periods into shorter time intervals. Bulk fuels (e.g. diesel, petrol, LPG, coal, etc) are often stockpiled on site, which may lead to differences between the fuel s purchase and subsequent use. It is important that data is captured for the time period (quantities and dates/times) when the fuel is used. This may require downstream metering (e.g. fuel pump readings taken when vehicles are filled, rather than relying on bulk fuel purchases which may be periodic or sporadic). Internal sub-metering can be used to describe a portion of a facility (e.g. buildings within a larger campus) or major energy systems. Routine adjustments and performance outputs via independent variables The identification of independent variables that affect a site is an important step for assessing sites using Options C or D. It is important to consider variables that may influence use across the measurement boundary, and these may include variables that are within the site s control as well as those that occur naturally. Examples of independent variables include: Weather including ambient temperature, humidity, rainfall, wind speed/direction Operating hours daily/weekly/seasonal operating schedules Occupancy staff, students, patients, shoppers, tenants, visitors, etc System loads/ activity levels heating/cooling requirements, temperature set points, work required from equipment, etc Input raw materials temperatures, purity, density, moisture content, etc Production types and amounts product quantities, volumes, weight, etc Where a site engages in multiple activities, data for each may be required in order to effectively model site energy use. With all independent variables it is important that the range of possible values is known. Non-routine adjustments via static factors Given the long-term approach taken when using Option C and D, the effects of non-routine adjustments may become significant and require once off or permanent adjustments. The identification and collection of data relating to static factors may assist with identifying or adjusting outlying data points, or with making a permanent step change in modelling forecast figures. Typical static factors include material and permanent changes to independent variables listed above, as well as: Change in product mix (e.g. plant stops making product A and start making product B) Equipment retrofits (either directly related to energy or not) Production schedule/shift changes Building changes (renovation, extension, facade changes, etc) Extreme and infrequent weather events (flood, fire, cyclone, etc) Effects of ECMs Interactive effects Since total facility energy consumption is captured within the measurement boundary, interactive effects between energy systems will be accounted for. 4.1

15 M&V design and planning Operating cycle When using Option C or D, the baseline period is 12 months or more, and post-retrofit period is typically 6 months or more Additionality Since total facility energy consumption is captured within the measurement boundary, additionality will not be an issue for the M&V project. 4.2 Prepare M&V plan The next step of the M&V process is to prepare an M&V plan which is based on the M&V design and the time, resources and budget necessary to complete the M&V project. Refer to the Planning Guide for further guidance on preparing an M&V plan. The table below outlines issues commonly found when conducting M&V on whole building projects and provides suggested approaches for addressing them in you M&V plan and when executing the M&V project. Table 3: Considerations, issues and suggested approach for whole building projects Consideration Issue Suggested Approach Modelling software and modelling/simulation skills Insufficient skill or familiarity with complex modelling tools Seek formal training on use of specific software packages. The software vendor is a useful place to start enquiries. OR Engage a suitably qualified professional to conduct the analysis Difficult sites to model Certain site types Consider variations to linear regression, including: Multivariable regression Non-linear curves Stepped regression Separate regression for each fuel type Consider alternatives to regression including bottom up energy use models using Option D. Validating regression models Understanding the process for validating regression models Refer to the Process Guide for further information. 4.2

16 12 Measurement and Verification Operational Guide 5 Data collection, modelling and analysis 5.1 Measure baseline data Measurement data sources, measurement tools and techniques The following provides general guidance on measurement and data collection: Conduct baseline measurement in line with the prepared M&V plan prior to implementing the project. Ensure appropriate records are kept including the placement of measuring equipment and take lots of photographs. Collect any associated data required for calculating baseline energy use or adjustments for independent variables. Measurement should consider period demand and, where applicable, measurement should be made during any and all relevant demand periods Conducting measurements Option C The following provides guidance on the approach and placement for conducting measurements specific to Option C: Energy use measurements will be conducted using permanent metering. Data for independent variables should also be collected using ongoing, robust and quality assured processes. Data may be collected electronically from control systems (e.g. BMS, SCADA systems), or be the result from manual counts or surveys. Meter calibration should be considered periodically. Option D Under Option D, measurements are taken only within the post-retrofit period. 5.2 Develop energy model and uncertainty This is the most critical step when using Option C or D, as we aim to develop an energy model that explains the baseline energy use as accurately as possible. Spreadsheets may be adequate for simple models however specialised regression analysis software may be required for models that involve multiple independent variables and the adjustment of weather data against target temperature balance points. Where possible it is a good idea to develop the energy model prior to implementing the ECM in case problems arise, that may require a delay or change or M&V Option from C to either A, B or D. When using Option D, data for relevant inputs described in are input into the building simulation program. The software produces a simulated energy output at either daily or hourly frequency. Separate models will need to be developed; one with the ECMs and one without. Uncertainty can be introduced into the energy model due to inaccuracies of measurement equipment and regression modelling errors. These inaccuracies need to be quantified as an overall uncertainty statement which includes a precision and confidence level. Refer to the Process Guide for further guidance on calculating and expressing uncertainty. 5.1

17 Data collection, modelling and analysis Implement ECM(s) Implement the ECM. Under Option C, this may have already occurred, which shouldn t be an issue if a successful energy model has been determined. Ensure that appropriate commissioning has been conducted to confirm that the ECMs are operating as expected. Under Option D, the new building is constructed or the ECM is implemented. 5.4 Measure post retrofit data Option C Conduct post-retrofit measurement in line with the prepared M&V plan using the same techniques as for the baseline (section 5.1). Collect any associated data required for calculating post-retrofit energy use or adjustments based on independent variables (e.g. monthly heating and/or cooling degree days). Confirm data integrity and completeness. Option D Post-retrofit data for energy use, independent variables and static factors is collected, allowing a suitable time to commission the building or bed in a new ECM. Measurements are taken only within the post-retrofit period. Energy use measurements will be conducted using permanent metering. Data for independent variables should also be collected using ongoing, robust and quality assured processes. Data may be collected electronically from control systems (e.g. BMS, SCADA systems), or be the result from manual counts or surveys. Meter calibration should be considered periodically. 5.5 Savings analysis and uncertainty Savings equations Option C The energy model previously developed is now used to forecast across the post-retrofit the business as usual consumption which would have occurred without the ECM. This is achieved by applying the various data inputs specified by the model for independent variables and static factors. Savings are then calculated by subtracting the business as usual forecast from the actual savings from the metered energy use. From an uncertainty point of view, billed energy data is considered 100% accurate. Other uncertainties will be dependent on the modelling uncertainty, plus those from data collection. The general equation for energy savings is: Energy Savings = (Baseline Energy Post-Retrofit Energy) ± Adjustments 5.3

18 14 Measurement and Verification Operational Guide Option D Calibrate the building simulation against post-retrofit energy use and conditions to calculate savings: Both developed simulation models (simulation models with and without ECMs) are recalibrated using data from the post-retrofit period for all key independent variables. The installed equipment and controls, internal heat loads and other characteristics are confirmed and adjusted. Changes are applied to both models as required, based on the concept of calibrating the model that contains the ECMs (i.e. as built model) so that it accurately reflects the metered energy use from the measurement boundary. It should confirm that it appropriately describes both the energy use and demand profile from monthly data. 12 months of post-retrofit data is used to calibrate the model. Actual weather data is also used. Once off or short term measurement of individual equipment may assist with appropriately calibrating the model for complex systems. The calibration error is determined at a monthly level using the as-built model and the actual energy data. Calibration errors are applied to the baseline model in order to incorporate uncertainty within savings. Savings are calculated in one of two ways: Actual savings = Actual energy use calibrated baseline model, OR Normalised savings = calibrated as-built model calibrated baseline model Actual savings are determined for the post-retrofit period, by applying the appropriate data to the baseline model. Normalised savings can be determined using any set of data for independent variables (e.g. 10 year average weather patterns), once the models are calibrated. The data is applied to both models Extrapolation Extrapolate the calculated savings for the measured period as required Uncertainty Estimate the savings uncertainty, based on the measurement approach, placement, impact of variables, length of measurement and equipment used. Refer to the Process Guide for further guidance on calculating and expressing uncertainty. 5.5

19 Finish 15 6 Finish 6.1 Reporting Prepare an outcomes report summarising the M&V exercise. Ensure any extrapolated savings are referred to as estimates, as the actual savings only apply to the measurement period. Energy uncertainty is expressed with the overall precision and confidence level. 6.2 Project close and savings persistence Under Option C, ongoing M&V can be conducted regularly as required by collecting independent variables and inserting them into the energy model. Typically under Option D, the first year of data forms the baseline of an ongoing Option C M&V approach. 6.1

20 16 Measurement and Verification Operational Guide 7 M&V Examples Both the IPMVP and this guide contain several worked example M&V projects. These are provided to assist readers with applying M&V concepts in real world situations, and to demonstrate the design and analytical components of successful M&V projects. 7.1 Examples from the IPMVP The table below lists the example M&V projects that can be found within the IPMVP. Table 4: Example M&V projects from the IPMVP M&V Project Name IPMVP Option Location Pump/Motor Efficiency Improvement A Volume 1: Appendix A A-2 Pump/Motor Demand Shifting B Volume 1: Appendix A A-2-1 Lighting fixture upgrade A Volume 1: Appendix A A-3 Lighting control A Volume 1: Appendix A A-3-1 Lighting new fixtures and dimming B Volume 1: Appendix A A-3-2 Compressed-Air Leakage Management B Volume 1: Appendix A A-4 Turbine/Generator Set Improvement B Volume 1: Appendix A A-5 Boiler Efficiency Improvement A Volume 1: Appendix A A-6 Multiple ECMs with metered baseline data C Volume 1: Appendix A A-7 Whole facility energy accounting relative to budget C Volume 1: Appendix A A-7-1 Multiple ECMs in a building without energy meters in the baseline period D Volume 1: Appendix A A-8 New building designed better than code D Volume 1: Appendix A A-9 Solar water heating test A Volume 3: Renewable Energy Direct measurement centralised solar hot water heater B Volume 3: Renewable Energy Indirect measurement residential solar hot water heater B & D Volume 3: Renewable Energy Building integrated photovoltaic system D Volume 3: Renewable Energy Solar Water Heating D Volume 3: Renewable Energy 7.1

21 M&V Examples Examples from this guide The table below lists the example M&V projects that can be found within this guide. Table 5: Example M&V projects from the M&V Operational Guide M&V Project Name IPMVP Option Location M&V design examples A, B, C, D Process: Appendix A Demand and cost avoidance calculation example Regression modelling and validity testing Lighting fixture replacement within an office tenancy Lighting fixture and control upgrade at a function centre Lighting fixture retrofit incorporating daylight control Pump retrofit and motor replacement Car park ventilation involving CO monitoring and variable speed drive on fans Replacement an inefficient gas boiler with a high efficiency one Upgrade freezer controls within a food processing plant Compressed air leak detection within a manufacturing site using sampling analysis Steam system leak detection within a food processing site using regression analysis Multiple ECMs involving compressed air and steam system optimisation, combined with lighting controls at a cannery Commercial building air conditioning central plant upgrade Evaluate performance efficiency of a newly installed cogeneration unit a a school Installation of a cogeneration plant at a hospital Use of solar hot water system on a housing estate n/a n/a A A B A B C B B B C C D C B Process: Appendix A Process: Appendix E Applications: Lighting Scenario A Applications: Lighting Scenario B Applications: Lighting Scenario C Applications: Motors, Pumps and Fans Scenario A Applications: Motors, Pumps and Fans Scenario B Applications: Heating, Ventilation and Cooling Scenario A Applications: Commercial and Industrial Refrigeration Scenario A Applications: Boilers, Steam and Compressed Air Scenario A Applications: Boilers, Steam and Compressed Air Scenario B Applications: Whole Buildings Scenario A Applications: Whole Buildings Scenario B Applications: Renewables and Cogeneration Scenario A Applications: Renewables and Cogeneration Scenario B Applications: Renewables and Cogeneration Scenario C 7.2

22 18 Measurement and Verification Operation Guide Appendix A: Example scenario A food processing plant The scenario below provides details of how Option C is used to measure and verify the savings from multiple ECMs across a manufacturing plant. A food processing plant in a regional town prepares and packages a range of tinned and snap frozen vegetable products grown in the local area. The plant operates year-round, however activity in a given month is seasonal based on the produce and consumer spending patterns. The plant consists of several production lines, some of which serve multiple products throughout the year. Broad beans, which are the plant s most popular product is highly seasonal and is only produced for 5 months of the year. The basic process involves food preparation (skinning, dehusking, etc), cooking, and either snap freezing and packaging, or canning. A centralised set of boilers provide hot water and steam for cooking and sterilisation, whilst a set of air compressors provide compressed air for operating much of the machinery. The site is a significant energy user and receives monthly invoices for electricity and natural gas. At present there is no sub-metering of energy use across the site. Following an energy audit, the Site Manager has gained approval to proceed with implementing several ECMs across the site. At first the low cost measures are to be implemented and once savings can be verified, capital will be available for more significant upgrades. The identified ECMs are mostly small to medium and include: 1. Adding a heat exchanger to capture waste heat from the air compressor system to pre-heat boiler feedwater. 2. Reducing compressed air and steam delivery pressures 3. Introducing occupancy based lighting controls for seldom used areas When combined together, the Site Manager estimates that the above ECMs will reduce site energy use by around 15%. Each ECM relates to a different energy system, and there is the potential to assess each individually. Monthly energy use is highly variable, which the Site Manager believes is due to changes in product output, however she cannot be sure. The Site Manager is only interested in determining combined savings, and does not have the funds available to invest in temporary metering. More importantly, the Site Manager seeks to develop an approach for conducting M&V that will apply into the future as more ECMs are implemented. For these reasons, an Option C approach is to be used to verify savings. Management are interested in savings to a confidence of 90%. Appendix A

23 Appendix A: Example scenario A food processing plant 19 Getting started Budget The potential savings (15% of site usage) from these projects are significant, however the site is has complex, interlinked manufacturing processes. Due to the difficulty of defining appropriate project measurement boundaries, combined with the fact that the Site Manager does not have funds for temporary metering, an Option C approach has been adopted which simply involves desktop analysis using existing data. It is anticipated that the analysis can be conducted and savings calculated within 12 hours. A nominal budget of $2,500 of time has been set aside. Measurement boundary The measurement boundary is the combination of electrical and natural gas supplies that supply the facility. Both fuels will be affected by the proposed ECMs. Although the Site Manager is only interested in overall savings, separate analysis will be conducted for electricity and natural gas in order to calculate cost savings. Identification of key parameters After working at the site for a number of years in various roles, the Site Manager has developed her opinion regarding the variables that may affect energy use. She hasn t conducted detailed analysis before to determine these, and so she plans on collecting data for the following variables and attempting various models to find the best combinations of data. The variables to be investigated include: Variable Metered electricity and natural gas use Data source Monthly electricity invoices Ambient air temperature (heating and cooling degree days) Daily figures for maximum and minimum air temperatures from the Bureau of Meteorology Production figures for various product lines, divided into: Canned Mixed Vegetables Broad beans Peas Other production (combined) Monthly site reports which include data compiled from shift-based production outputs The plant operates continuously, and so operating hours will not be considered. However some products are highly seasonal, which will be reflected in the monthly production volumes. Timing Baseline data is available for the last 3 years. The ECMs are due to be implemented during September 2010, and so the period just prior has been chosen. The baseline period will consist of the 12 months between September 2009 and August Interactive effects There are no interactive effects due to the choice of a facility based measurement boundary. Appendix A

24 20 Measurement and Verification Operational Guide Summary of M&V plan The key elements of the project s M&V plan in summary are: Item Project Summary Plan Implementation of the following ECMs on central service equipment: waste heat capture from the air compressor system to pre-heat boiler feedwater pressure reduction strategy for compressed air and steam systems occupancy based control of lighting in seldom used areas Required Outcome To confirm savings in excess of 15% with a 90% confidence level in order to convince management to invest in more capital intensive energy efficiency projects. Budget $2,500 (1 day of desktop analysis) M&V Option C Whole Facility Measurement Boundary Total incoming electrical and natural gas supplies being fed to the BeanCo Central West Processing Plant, located in 23 Rabie Road, Parkes, NSW. Key Measurement Parameters Other Parameters to consider Data for the following key parameters will be collected and modelled in order to determine the most relevant factors affecting energy use: Metered electricity and natural gas use Ambient temperature (in the form of heating/cooling degree days) Monthly production figures for processed goods Canned mixed vegetables Broad beans Peas Other food products None identified Potential interactive effects None identified Approach for conducting measurement and collecting data Energy use data will be collected directly from energy invoices. Ambient temperature data will be sourced for the nearest weather station from the Bureau of Meteorology. Production data will be obtained from monthly reports, which are compiled from daily shift data. Measurement equipment required Existing revenue meter and energy invoice data Measurement period 12 month base year period, 1 month project implementation, and 6+ month post-retrofit measurement. Appendix A

25 Appendix A: Example scenario A food processing plant 21 Item Approach for calculating results Plan Separate energy models for electricity and natural gas will be developed using regression analysis to explain baseline energy use in regard to relevant independent variables. As the relationship between independent variables and energy use is unknown and potentially complex, several energy models will be tested to identify the variable (or combination of variables) that best describe energy use for each fuel type. A preferred energy model will be selected for each fuel type, based on highest R 2 value, satisfying regression validity tests and having the smallest standard error. Once the ECM has been implemented, post-retrofit data for these variables will be entered into the fuel specific models in order to forecast the adjusted baseline or business-as-usual energy use. Savings are calculated by taking the difference in the adjusted baseline and actual energy use data on a monthly basis. Cost savings are calculated by applying the following energy rates: Natural gas - $7.30 / GJ Electricity - $0.18 / kwh Baseline model development The decision was made to conduct the modelling using Microsoft Excel. As it is not clear which variables will best explain monthly energy use, several scenarios will be modelled and compared. A spreadsheet was developed that contained the following elements: Monthly baseline historical data for the key energy use parameters in independent variables Daily weather maximum and minimum temperatures for the local weather station HDDs and CDDs that have been derived from ambient temperature data, as described within Section within the Process Guide. The baseline site data is summarised below: Month Energy use Production Data (tonnes) Total Electricity (kwh) Natural Gas (GJ) Canned Mixed Vegetables Broad beans Peas Other Production Production (tonnes) B1 537,825 3, B2 496,767 2, B3 696,798 3, B4 498,189 2, B5 653,955 2, B6 1,069,023 4, ,459 B7 712,298 3, B8 898,471 4, ,992 Appendix A

26 22 Measurement and Verification Operational Guide Month Energy use Production Data (tonnes) Total Production Peas (tonnes) Electricity (kwh) Natural Gas (GJ) Canned Mixed Vegetables Broad beans Other Production B9 882,253 5, ,024 B10 466,539 3, B11 451,487 4, B12 488,450 2, In addition to the data above, weather data was collected from the Bureau of Meteorology for the nearest weather station. From the average daily temperatures, values for Heating Degree Days (HDDs) and Cooling Degree Days (CDDs) were computed. Initial values for HDD and CDD balance points were chosen, and were added into the analysis spreadsheet as reference values in linked cells. The balance points were adjusted as part of tuning various energy models so that the R 2 value was maximised for each energy model. The LINEST() function is used to conduct multivariable linear regression on a dataset to produce a function describing y in terms of x 1 to n in the form: where: y = b + m 1 x 1 + m 2 x m n x n y = energy use b = baseload energy use (without influence of any variables) also known as y-intercept x 1 to n = data for independent variables 1 to n m 1 to n = coefficients determined by the function that correspond to each x variable It accepts an array of y-values (i.e. energy use) and corresponding arrays of x-values (data for independent variables) and produces various outputs as shown below: Coefficients and standard errors for the baseload energy use (b) and each independent variable (m 1 to n ) Coefficient of determination (R 2 ), and Adjusted R 2 which describe the appropriateness of fit for the regression model Standard error for y Developing the energy model for electricity Seven different regression models were calculated, each incorporating a different combination of parameters as shown in the table below. Appendix A

27 Appendix A: Example scenario A food processing plant 23 Scenario Electricity (kwh) Canned Mixed Vegetables Production(tonnes) Broad Beans Total Production (tonnes) Ambient temperature Peas Other HDDs CDDs 1 Yes Yes Yes Yes Yes - 3 Yes Yes - Yes 4 Yes Yes Yes Yes Yes 5 Yes Yes Yes Yes Yes Yes Yes - 7 Yes Yes Yes Yes - - Yes - The combinations above were selected as the analysis proceeded. As each model was developed, the t-stats for each coefficient were examined to see which variables were providing a strong correlation. Any coefficient with a t-stat value less than 2 was further examined in subsequent models. In addition, changes in R 2 provided a guide as to how each model reacted to inclusion of new variables, based on a base model involving total production. Refer to Help within Microsoft Excel for assistance with using the LINEST() function. An example of the output using the input data above for Scenario 6 is shown below: coefficients R 2 standard error for y standard errors for coefficients The output above equates to the following: Energy use = 334, t-stats calculated separately x (HDD) x (Canned Mixed Vegetables (tonnes)) + 1, x (Broad Beans (tonnes)) The R 2 value is and the standard error is 55, Each model was generated in turn using the LINEST() function. Appendix A

28 24 Measurement and Verification Operational Guide Validating the regression outputs As described in Section 6.5.1, each model must be validated to confirm that it can be used. This involves equating some additional attributes and confirming their validity. The attributes to be reviewed are: Attribute Description and Validity Test R 2 Must be 0.75 or higher. Output from LINEST() function Adjusted R 2 t-statistic for each coefficient To be evaluated when multiple independent variables are used. Values should be 0.75 or higher. Calculated as follows: (n 1) Adjusted R 2 = 1 (1 R 2 ) (n k 1) Where: R 2 = the value calculated from LINEST() n = number of data points in sample size k = number of coefficients within the regression model Determine the validity of each x coefficient. Values must be greater than 2. Calculated as follows: Each t-stat is calculated as follows: t Xn = coefficient Xn standard error Xn Absolute mean bias (MBE) Determines the overall bias in the regression estimate. MBE values should be < 0.005%. Calculated as follows: MBE = (modelled n actual n ) n Where: n = the number of samples modelled n = predicted value for sample item n actual n = actual value for sample item n Monthly Mean Error CV(RMSE) Related to the standard error of the model. CV(RMSE) values should be < Calculated as follows: CV(RMSE) = standard error for y average(actual values) Appendix A

29 Appendix A: Example scenario A food processing plant 25 In our example above: Adjusted R 2 (12 1) = 1 ( ) (12 3 1) = For our three coefficients in the example above: t canned mixed vegetables = = 4.07 t broad beans = 1, = t HDDs = = 2.05 MBE is calculated by first determining the modelled energy values by using the data for independent variables. Modelled values are then subtracted from actual values to calculate the difference for each measurement. These residuals are summed and divided by the number of data points to determine the MBE, as follows: Month Actual energy Example Scenario 6 modelled energy Residual (modelled actual) B1 537, ,614 15,789 B2 496, ,050 8,283 B3 696, ,419-46,379 B4 498, ,030 45,841 B5 653, ,301 77,346 B6 1,069,023 1,005,560-63,463 B7 712, ,574-50,724 B8 898, ,046 43,575 B9 882, ,876 9,623 B10 466, ,269-14,270 B11 451, ,067 26,580 B12 488, ,249-52,201 Sum x Average 654, And, MBE is calculated as follows: MBE = = CV(RMSE) is calculated for the example above as follows: CV(RMSE) = 53, , = Appendix A

30 26 Measurement and Verification Operational Guide Validity tests for each of the seven models were conducted. The results are shown in the table below. Conditional formatting is used to highlight test results. Each value is coloured green if it passes the validity test and red if it fails. Scenarios 5 and 6 pass all the validity tests. Scenario 6 was selected from the table above as the preferred scenario as it had the highest R 2 and Adjusted R 2 values and the lowest standard error. Summarising the electricity model The output from the LINEST() function for Scenario 6 is shown below. From the above the energy model describes monthly energy use as: Electricity use(kwh) = 334,633 + ( HDDs) + ( Canned Mixed Vegetables) + (1, Broad Beans) And the standard error is ± 53, kwh Calculating electricity savings The ECM was implemented in September Following a number of months, the savings were calculated for the ECM by forecasting the adjusted baseline by applying post-retrofit data to the baseline energy model. This is shown in the table below. Appendix A